Work

Guardian-in-the-Loop: Professional Oversight Beyond Human-in-the-Loop

Year: 2026

Type: article

Abstract: Guardian-in-the-Loop is a professional oversight model that improves on “human-in-the-loop” by making accountability explicit, enforceable, and evidence-bound. Human-in-the-loop can fail when it is treated as a vague checkpoint—someone reviews an output, but the workflow does not specify who is qualified, what standards must be applied, or what happens when evidence is missing. Guardian-in-the-Loop defines the responsible domain professional (e.g., legal reviewer, clinician/compliance lead, audit/risk lead, archivist/curator/linguist) with authority to approve, pause, or reject claims based on evidence. This matters across two arenas: human preservation (law, healthcare, finance) and cultural preservation (language, translation, writing systems, archives, heritage records). The paper connects this to documented risks from model hallucinations, including legal incidents involving fabricated citations. An open-source implementation is provided through falsifiability sheets, a two-rater review process, and a traffic-light rubric (Green/Yellow/Red) so claims can be scored as supported, uncertain, or not supported based on evidence. Related Zenodo records are linked for reproducible templates and worked examples.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.18489953

Author: Grasa, Michael

Institution: George Mason University

Language: English

Topic: Ethics and Social Impacts of AI

Subfield: Safety Research

Field: Social Sciences

Domain: Social Sciences

SDG: Peace, justice, and strong institutions

Open Access status: green

Funder: George Mason University